CN-117584409-B - Injection product quality prediction method and system based on XGBoost-LSTM integrated model

CN117584409BCN 117584409 BCN117584409 BCN 117584409BCN-117584409-B

Abstract

The invention discloses an injection product quality prediction method and system based on XGBoost-LSTM integrated model, wherein the method comprises the following steps of S1, designing a pouring system and a cooling system according to injection product requirements, determining key injection process parameters, performing model flow analysis and simulation actual production, S2, collecting Moldflow simulation analysis data, S3, preprocessing a simulation data set, dividing the simulation data set into a training set and a testing set, S4, training the XGBoost-LSTM integrated model through the training set, and S5, verifying the prediction accuracy of the XGBoost-LSTM integrated model through the testing set. The system comprises a model flow analysis module, a simulation data acquisition module, a data preprocessing module, a model construction module, a model training module and a model test module. According to the invention, the product quality is accurately predicted according to the set injection molding parameters, the trial-and-error cost is reduced, and the production efficiency is improved.

Inventors

MO SHENGYONG

Assignees

峰睿新科智能有限公司

Dates

Publication Date: 20260505
Application Date: 20231120

Claims (6)

1. The injection product quality prediction method based on XGBoost-LSTM integrated model is characterized by comprising the following steps of: S1, designing a pouring system and a cooling system by utilizing Moldflow software according to the requirements of injection products, determining key injection molding process parameters based on orthogonal test design, and performing die flow analysis to simulate actual production; S2, acquiring Moldflow simulation analysis data to obtain a simulation data set; s3, preprocessing the simulation data set, and dividing the preprocessed simulation data set into a training set and a testing set; S4, constructing XGBoost-LSTM integrated models, and training the XGBoost-LSTM integrated models through a training set to obtain a regression model for quality prediction of injection products; S5, verifying the prediction precision of XGBoost-LSTM integrated model through a test set; in step S4, the XGBoost-LSTM integrated model comprises an LSTM sub-model, a XGBoost sub-model and a meta-model, wherein the XGBoost-LSTM integrated model is trained through a training set, and the method comprises the following steps of S41, inputting real-time modeling state data in the training set into the LSTM sub-model to obtain a first predicted value, S42, inputting injection molding process parameter data and product quality data in the training set into the XGBoost sub-model to obtain a second predicted value, S413, inputting the first predicted value obtained in step S41 and the second predicted value obtained in step S42 into the meta-model, wherein the meta-model adopts a weighted combination of the predicted values of all the sub-models as a final predicted value, and the meta-model adopts a calculation formula of the weighted combination of the predicted values of all the sub-models as follows: , wherein, The weight of the LSTM sub-model predictive value, Weights for XGBoost submodel predictors, As a result of the final predicted value, As a result of the first prediction value, Is the second predicted value.
2. The method for predicting the quality of an injection molding product based on XGBoost-LSTM integrated model as set forth in claim 1, wherein the key injection molding process parameters include injection time, dwell time, cool time, injection pressure, dwell pressure, material temperature, nozzle temperature, and mold temperature.
3. The method for predicting the quality of an injection molding product based on XGBoost-LSTM integrated model as set forth in claim 1, wherein the simulation dataset comprises a process parameter set value, product quality data and real-time mold flow state data, wherein the product quality data comprises mold filling time, a welding line, flow front temperature, volume shrinkage and warp deformation, and the real-time mold flow state data comprises pressure, mold locking force, volume and temperature at an injection position.
4. The method for predicting the quality of an injection molding product based on XGBoost-LSTM integrated model as set forth in claim 3, wherein the preprocessing of the simulation dataset comprises the steps of: S31, eliminating abnormal values in the product quality data; s32, carrying out normalization processing on the simulation data set; s33, carrying out data segmentation and reconstruction on the real-time modeling state data.
5. The method for predicting the quality of an injection molding product based on XGBoost-LSTM integrated model as set forth in claim 4, wherein the key of data segmentation and reconstruction of real-time modeling state data is that sequence segments are divided by adopting a sliding window mode with time steps as intervals, and are recombined into new feature input.
6. An injection product quality prediction system based on XGBoost-LSTM integration model, adopting the injection product quality prediction method based on XGBoost-LSTM integration model as set forth in any one of claims 1-5, comprising: The die flow analysis module is used for designing a pouring system and a cooling system by utilizing Moldflow software, determining key injection molding process parameters based on orthogonal test design, and simulating actual production by die flow analysis; the simulation data acquisition module is used for acquiring Moldflow simulation analysis data to obtain a simulation data set; The data preprocessing module is used for preprocessing the simulation data set and dividing the preprocessed simulation data set into a training set and a testing set; The model construction module is used for constructing XGBoost-LSTM integrated models, wherein the XGBoost-LSTM integrated models comprise LSTM sub-models, XGBoost sub-models and meta-models; The model training module is used for inputting a training set into the XGBoost-LSTM integrated model to train to obtain a regression model for the quality prediction of the injection product, and And the model test module is used for verifying XGBoost-LSTM integrated model prediction accuracy through the test set.

Description

Injection product quality prediction method and system based on XGBoost-LSTM integrated model Technical Field The invention relates to the technical field of injection molding production, in particular to an injection molding product quality prediction method and system based on XGBoost-LSTM integrated model. Background In the injection molding industry, quality control of injection molded product pieces is concerned, and is one of core technologies for controlling production cost and improving market competition advantages of enterprises. The traditional injection molding product quality control method at present mainly comprises the offline detection of various quality indexes after production, so that the cost is huge in time and capital consumption, and the comprehensive and rapid detection of the product quality cannot be ensured, so that the overall production level and economic benefit of enterprises are difficult to improve. In recent years, with the rapid development of artificial intelligence technologies such as machine learning and deep learning, a quality prediction method based on data driving is continuously developed and gradually applied to various industrial fields. At present, machine learning algorithms based on a support vector machine, response surface analysis, an artificial neural network, a random forest and the like are applied to quality prediction, but the problems that time sequence information in industrial data is not fully utilized, the precision of a single prediction model is unreliable and the like still exist at present, and difficulties and challenges are brought to stable and efficient production of enterprises. Disclosure of Invention Aiming at the defects in the prior art, the invention provides the injection molding product quality prediction method and the injection molding product quality prediction system based on XGBoost-LSTM integrated model, which can quickly establish a black box model in the injection molding production process, accurately predict the product quality according to the set injection molding parameters, effectively reduce the production trial-and-error cost of enterprises and improve the production efficiency. In order to solve the technical problems, the technical scheme of the invention is as follows: a quality prediction method of injection products based on XGBoost-LSTM integrated model comprises the following steps: S1, designing a pouring system and a cooling system by utilizing Moldflow software according to the requirements of injection products, determining key injection molding process parameters based on orthogonal test design, and performing die flow analysis to simulate actual production; S2, acquiring Moldflow simulation analysis data to obtain a simulation data set; s3, preprocessing the simulation data set, and dividing the preprocessed simulation data set into a training set and a testing set; S4, constructing XGBoost-LSTM integrated models, and training the XGBoost-LSTM integrated models through a training set to obtain a regression model for quality prediction of injection products; S5, verifying EXGBoost-LSTM integrated model prediction accuracy through the test set. As a preferred option, the key injection molding process parameters include injection time, dwell time, cool time, injection pressure, dwell pressure, material temperature, nozzle temperature, and mold temperature. As a preferable scheme, the simulation data set comprises a technological parameter set value, product quality data and real-time molding state data, wherein the product quality data comprises mold filling time, a welding line, flow front temperature, volume shrinkage and warp deformation, and the real-time molding state data comprises pressure, mold locking force, volume and temperature at an injection position. As a preferred solution, the preprocessing of the simulation data set includes the following steps: S31, eliminating abnormal values in the product quality data; s32, carrying out normalization processing on the simulation data set; s33, carrying out data segmentation and reconstruction on the real-time modeling state data. The key of the data segmentation and reconstruction of the real-time analog state data is that T time steps are selected as intervals, sequence segments are divided by adopting a sliding window mode, and the sequence segments are recombined into new characteristic input. As a preferred scheme, the XGBoost-LSTM integrated model comprises an LSTM sub-model, a XGBoost sub-model and a meta-model. As a preferable scheme, the XGBoost-LSTM integrated model is trained through a training set, and the method comprises the following steps of: s41, inputting real-time model state data in a training set into an LSTM sub-model to obtain a first predicted value; s42, inputting the injection molding process parameter data and the product quality data in the training set into XGBoost submodels to obtain a second predicted value; s43, inputt